00 production rules

Production Rules — wod.wiki Explainer

Scenes are rendered individually and clipped together. These rules are what makes seven separate renders feel like one film.

Canvas & framing

• Master: 16:9, 1920×1080. Landing-page embed may crop to a shorter player, never reframe scenes.
• Total assembled runtime: ≈1:33 (90-second class). Per-scene durations are locked in the outline timeline; a narration re-take that changes a scene's length by more than a second means re-timing that scene's animation spec, not stretching the edit.
• Board background = app page background (see UI guide). No vignettes, no paper texture, no drop shadows on the board itself.
• Safe margin 96px on all sides; no text, chip, or annotation crosses it.
• Camera: locked-off per scene; the only moves allowed are slow push-ins/pull-backs on the board's plane. One move per scene, except Scene 3 (the statement zoom), which is granted two. No pans across scenes, no 3D moves, no handheld shake.

Transitions (identical everywhere)

• Scene 2 → 3 → 4 is one continuous shot rendered in three pieces: the block may not move, resize, or restyle across the seams; Scene 4's first frame must equal Scene 3's last frame.
• All other cuts are hard cuts on a settled frame (nothing mid-motion at a seam), except Scene 6 → 7 (pull-back reveal) as written in the scene files.
• Clicks: the CHAR-3 ripple + click sound is the only "UI transition effect." No swooshes, sparkles, or screen flashes anywhere.

Annotation system

• Marker annotations (CHAR-2) are for emphasis only: underlines, one bracket set in Scene 3, short phrases (≤5 words). Max 3 per scene. Marker ink never labels UI permanently — if a label must persist, it's app text (Inter/JetBrains Mono).
• Arrows point; they never cross the Athlete or the timer digits.

Audio

• One narrator: Sergei (see characters.md). Same mic, room, distance, and gain in all scenes; record in one session where possible, and use the logged setup for any re-take. VO is mixed over a single shared music bed — same track, same level (−18 dB under VO) for the whole film; music never swells except a slight lift under Scene 7's end card.
• Sound effects are the app's own: the timer bell is the actual bell sound from the build; the click is one recorded click reused everywhere; keystrokes in Scene 2 are one sample set. Borrowed stock SFX are a consistency break.
• Silence is allowed: Scene 1's final beat and Scene 6's settle use it on purpose.

Files & naming

• Renders: scene-0N-<slug>-v<k>.mp4 beside the scene's .md file in scenes/.
• Masters + the assembled cut live in this folder: explainer-master-v<k>.mp4.
• Any change to a canonical asset (characters, UI guide, dataset T-1) bumps a note in that file's header and triggers a re-check of every scene that uses it.

Per-render consistency checklist (run before accepting any scene)

• Athlete proportions, outfit, and pose match characters.md
• Any workout text parses under the real grammar; numbers match dataset T-1
• UI screens traced from the current build (Inter / JetBrains Mono, accent sampled)
• Clicks use the standard ripple + click; bell is the app's bell
• ≤3 marker annotations, correct ink color, ≤5 words each
• One camera move (two for Scene 3); seams land on settled frames
• VO is the locked narrator voice/settings; music bed level unchanged
• Scene's first/last frames match its neighbors' continuity notes